Test of HA in proxmox VE 4.0 beta 1 ISO

[realter]

Hi guys

,

I tried to use HA with iSCSI in a cluster of 3 nodes and have a VM in a group called "test" (no restrited, no failback) with redundant network/multicast.

My question is "are there any configuration to do in proxmox's interface and in proxmox's machines?" because i tried to migrate VM by unplug the supply and nothing happen but i got the following "Error Connection error 595: Connection timed out".

Best regards

 

[t.lamprecht]

Hello.
After unplugging you have to wait at least 60 seconds, as that is the time the watchdog triggers the self-fencing and the remaining cluster nodes can lock the lost members agent_lock.

On the HA Tab in Datacenter you see which node is the current master, look in it's log (Node->Syslog) to follow the fencing/migration process.

 

[realter]

Hello.
After unplugging you have to wait at least 60 seconds, as that is the time the watchdog triggers the self-fencing and the remaining cluster nodes can lock the lost members agent_lock.

On the HA Tab in Datacenter you see which node is the current master, look in it's log (Node->Syslog) to follow the fencing/migration process.

Hello t.lamprecht,

Thank you for your reply! It helps to know that!

I tried it again and about 2 minutes later, it worked.

Best regards.

 

[realter]

Is it important to choose a specific disk cache for HA with iSCSI?


Best regards.

 

[t.lamprecht]

Sorry, made a mistake. The locks have a lifetime of two minutes, so that's the (max) amount to wait and so it makes sense that you had to wait that time and not the 60s.

AFAIK, write through and directs sync are the safer, albeit also the slower ones.
Maybe someone else has more experience with it.

 

[mir]

Cashing is controlled by the iSCSI target so choosing nocache is perfectly safe.